(1991), Br. J. Cancer (1991),
Br. J. Cancer
'."
3-9 63, XIV, 3 Suppi. XIV, 63, Suppl.
1991 Macmillan Press Ltd., 1991 Ltd.,
9
Data analysis of the Second International Workshop Cancer Antigens
on
Small Cell Lung
W.R. Gilks', L. Oldfield' & P. Wild2 'Medical Research Council Biostatistics Unit, Fair View Lodge, 5 Shaftesbury Road, Cambridge CB2 2BW, UK; 2Departement Environnement Chimique Service Epidemiologie, Avenue de Bourgogne, B.P. 27, 54501 Vandoeuvre Cedex, France. Summary Methods of data collection for the 2nd Small Cell Lung Cancer Workshop are described, and data reliability is reviewed. The method of cluster analysis of the workshop antibodies is described and discussed. Of the 27,1 11 results submitted 20,705 were judged to be reliable for analysis and 13,802 of these came from immunohistology experiments. Data derived from immunocytochemistry experiments were somewhat less reproducible than flow cytometry, immunohistology and ELISA experiments. The cluster analysis was developed from methods employed in the leucocyte antigens workshops. Several checks on the methods of cluster analysis and the transformation of data did not substantially alter the final groupings. The workshop confirms that, although there are some methodological difficulties, the cluster analysis can successfully be applied to data derived largely from immunohistology, and thus has applicability to other tumour types.
The main purpose of the data analysis of the Second International Workshop on Small Cell Lung Cancer (SCLC) Antigens was to cluster Workshop monoclonal antibodies into groups having similar patterns of reactivity. We report here on the methods of data collection and analysis and also present some summary results. More detailed reporting and discussion of results can be found elsewhere in this volume. Methods and results Data collection For the First SCLC Workshop, data were collected via paper forms. Following our initiative in the 4th Leucocyte Antigens Workshop (Gilks et al., 1989), participants in the Second Workshop were requested to submit their data on computer diskettes. To support this, data entry programs for use on IBM-PC microcomputers (or compatibles) were specially written, and sent to all participants. These programs allowed participants to enter, check, edit and if necessary delete their experimental data. Participants were also provided with paper forms to assist in the transfer of data from laboratory to computer. Three data entry programs were written: one for flow cytometry or fluorescence microscopy data; another for immuno-cytochemistry data, and the third for immunohistology data. The flow cytometry/fluorescence microscopy data entry program requested the following information for each experiment: target cell type (from a list of 20 normal and tumour cell types); species of secondary antibody (goat, rabbit, rat or sheep); and type of experiment (flow cytometry or fluorescence microscopy). For each antibody tested in each experiment, the data entry program requested the percentage of cells reacting (for flow cytometry, the percentage of cells fluorescing more intensely than a negative gate). Below, we refer to this as the amount of reactivity. The immunocytochemistry data entry program requested the following information for each experiment: target cell type (from the list referred to above); species of secondary antibody (as above); method of staining (indirect immuno-
peroxidase, PAP, APAAP, indirect immunofluorescence, immunogold or ABC); and method of fixation (none, acetone or formalin). For each antibody tested in the experiment the following were requested: the strength of staining (negative,
Correspondence: W.R. Gilks.
weak or strong); the amount of staining (0%, 1-25%, 26-75% or >75% of cells stained); and the location of staining (nucleus, cytoplasm, surface, or a mixture of these). The immunohistology data entry program requested the following information for each experiment: the type of tissue (from a list of 23 tissues); the species of secondary antibody (as above); the method of staining (as above); and the method of fixation (frozen - no fixation, frozen - brief acetone, formalin, or Bouin's). Also, the types of cell identified in the tissue were also requested (from a common list of 39 normal and tumour cell types). Then, for each cell type identifiable in the tissue, and for each antibody, the strength of staining (as above) and the amount of staining (as above) were
requested.
Where appropriate, each list of possible answers to each question included an 'other' option. Thus, for example, if a target cell used in a flow cytometry experiment was not included in the preset list of 20, the participant would be able to select the 'other' option and then record the full details. The items of data requested for the Second Workshop closely followed those of the First Workshop (Gilks et al., 1988), the major differences in the present workshop being that, for flow cytometry/fluorescence microscopy experiments no strength data were requested; for immunocytochemistry experiments the amount of staining was requested in the form of percentage bands and the location of staining was also
requested.
Collection of data via the data entry programs had the
advantage that participants were able to check and print out their own computerised data at source. It also allowed for rapid central data assimilation, shortening the time needed to prepare interim statistical analyses for feedback to participants. In the case of immunohistology, the data entry
program provided another important advantage. We found in the 4th Leucocyte Antigens Workshop that participants were reluctant to record negative reactions on all combinations of cell-type and antibody where they occurred, simply because of the multiplicity of such combinations in any immunohistology experiment. Instead, some participants recorded only reactions which were positive. This presented difficulties in analysing the data, as it was sometimes difficult to discern whether a cell-type did not react with any of the antibodies in the panel, or whether the cell-type was not present or not identified in the tissue sections. In the present Workshop, the immunohistology data entry program removed this problem by first asking the participant to state the cell-types identified in the tissue (as described above), and then forcing the participant to provide data for all these cell-types against each antibody tested. Clearly, enforcing such rigour in data recording would have been impossible using only paper data forms.
4
W.R. GILKS et al.
Data received For present purposes we define an experiment as the simultaneous testing of several antibodies. We define an individual experimental result as the data pertaining to one antibody in one experiment against one cell type. Table I summarises the experimental results recevied. The cell sonicate ELISA data was unsolicited, but was provided by one laboratory via the immunocytochemistry data entry program. These ELISA data derived from cytospins of cell lines which were grown in culture medium, harvested, washed, sonicated until 50-75% of the cells were broken, then used to coat plates in the presence of dilute glutaraldehyde, and then incubated with the primary antibodies, after which
peroxidase-conjugated bridging antibody and substrate were added and absorbances read at 492 nm. Several experiments did not include all Workshop antibodies: indeed, one laboratory included only four antibodies in each histology experiment. The amount of work done is more accurately reflected in the column headed 'Results contributed' in Table I. The numbers of results are roughly comparable with those of the First Workshop (Gilks et al., 1988). Data exclusions To ensure that data for analyses were reasonably reliable, experiments were excluded from analyses if negative controls in the antibody panel were found to react positively, or if there were large differences in reactivity amongst replicate antibodies (i.e. antibodies replicated in the Workshop panel). Thus the following groups of Workshop antibodies were used to determine data exclusions:
Antibody group Negative controls Replicate group 1 Replicate group 2 Replicate group 3
Workshop numbers 01,32,51,72 11,53 31,41,58,82 03,28,56,71
Antibody name PBS + 5% FCS MOC-31 NCC-LU-246
CAM5.2
The exclusion criteria were formulated as follows: (a) Flow cytometry: Exclude the experiment if either (i) > 10% amount of reactivity was recorded on any negative control or (ii) within two or more of the replicate groups, the amounts of reactivity on any pair of antibodies differed by more than 2.6 on the logistic scale. (The logistic transformation transforms a percentage y to the value lo&[y/ (100-y)]: this has the effect of stretching the ends of the percentage scale relative to the middle. For example, two replicate antibodies scoring 1% and 12% yield a difference of 2.6 on the logistic scale, as do scores of 12% and 65%) (b) Immunocytochemistry, Immunohistology and Cell Sonicate ELISA: Exclude the experiment if either (i) >0% amount of staining was recorded on any negative control antibody or (ii) within two or more of the replicate groups, 0% amount of staining was recorded on one antibody but not on another, or 1-25% was recorded on one and 76-100% was recorded on another. Table I summarises the experiments excluded, and the experimental results remaining.
Data reliability Tables II-VI demonstrate data reliability, by way of comparisons on five duplicate pairs of antibodies after making the exclusions detailed above. These five duplicate pairs were as follows, by Workshop number: 11 and 53 (both MOC-3 1); 31 and 41 (both NCC-LU-246); 58 and 82 (again both NCC-LU-246); 3 and 28 (both CAM5.2); and 56 and 71 (again both CAMS.2). To introduce the methodology, Table II compares antibodies 11 and 53 in Flow Cytometry experiments. To facilitate comparisons with Immunocytochemistry and Immunohistology data (Tables IV and V), amounts of reactivity in Table II have been grouped into ranges of percentages. Thus, for example, out of nine flow cytometry experiments in which antibody 11 scored in the range 1-25%, antibody 53 scored outside that range only once. As expected for a duplicate pair of antibodies, the results in Table II are concentrated on the
diagonal.
To summarise results on duplicate antibodies in flow cytometry experiments, Table III combines the results from Table II with similar results on the remaining four duplicate pairs of antibodies. Thus the 70 results in the second row and second column of Table III include the eight results from the corresponding cell of Table II. As expected, the results concentrate strongly on the diagonal of Table III, although the number of off-diagonal results (indicating lack of experimental precision) is perhaps surprising. In Table IVa, the reliability of the amount of staining in immunocytochemistry experiments is demonstrated as for flow cytometry experiments in Table III. Table IVa suggests that immunocytochemistry experiments are somewhat less reliable than flow cytometry experiments; (recall also from Table I that a high proportion of immunocytochemistry experiments were excluded before compiling Table IV). Table IVb presents a similar analysis based on the strength of staining in immunocytochemistry experiments, and demonstrates that strength of staining is no more or less reliable than amount of staining. Table Va demonstrates the reliability of the amount of staining in immunohistology experiments. It has been constructed similarly to Tables III and IVa, except that each result tabulated represents the reactions of a duplicate pair of antibodies in one cell-type in one experiment. Table Va shows that a negative reaction on one antibody is generally accompanied by a negative reaction on its duplicate. A positive (76-100%) reaction is about as reliable as for flow cytometry experiments. However, an intermediate (1-75%) reaction on one antibody bears no relationship at all with the reaction on its duplicate. Similar conclusions can be drawn from Table Vb, regarding the reliability of strength of staining data from immunohistology experiments. Table II Comparison of antibodies 11 and 53 (both MOC-31) in Flow Cytometry experiments Amount of reactivity Amount of reactivity on antibody 53 on antibody 11 0% 1-25% 26-75% 76-100% 0% 4 1 1 0 1-25% 0 8 1 0 26-75% 0 3 0 1 76-100% 0 0 0 9 The numbers in the body of the table are numbers of experiments.
Table I Data contributed to the 2nd SCLC Workshop Laboratories Experiments Results Experiments Results Experimental type contributing contributed contributed excluded remaining Flow cytometrya 7 49 3,643 6 3,006 Immunocytochemistry 4 34 2,329 11 1,405 Immunohistology 12 196 18,646 13 13,802 Cell sonicate ELISAb 1 26 2,493 0 2,493 All 305 20c 27,111 30 20,705 aNo Fluorescence Microscopy data were contributed. bExperimental data of this type wre not solicited. cThree laboratories contributed data from more than one type of experiment.
DATA ANALYSIS
Table III Reproducibility in Flow Cytometry experiments Amount of reactivity on second mab of pair Amount of reactivity 1-25% 26-75% 76-100% in first mab of pair 0% 1 2 2 21 0% 0 1 70 6 1-25% 2 3 1 15 26-75% 23 1 0 0 76-100% The table totals numbers of results on five pairs of duplicate monoclonals: (workshop numbers 11 and 53; 31 and 41; 58 and 82; 3 and 28; and 56 and 71). Table IV Reproducibility in Immunocytochemistry experiments (a) Amount of staining Amount of staining in 2nd mab of pair in 1st mab of pair 1-25% 26-75% 76-100% 0% 1 0 2 25 0% 3 3 1-25% 4 0 2 3 2 0 26-75% 21 2 0 0 76-100% Strength of staining in 2nd mab ofpair (b) Strength of staining Strong Weak in Ist mab of pair Negative 0 3 25 Negative 6 5 2 Weak 5 21 1 Strong The table totals numbers of results on the five duplicate pairs of monoclonals in Table II. Table V Reproducibility in Immunocytochemistry experiments Amount of staining in 2nd mab of pair (a) Amount of staining 1-25% 26-75% 76-100% 0% in Ist mab of pair 4 10 3 379 0% 1 0 9 6 1-25% 4 8 7 6 26-75% 8 85 1 1 76-100% Strength of staining in 2nd mab of pair (b) Strength of staining Weak in Ist mab of pair Strong Negative 6 10 379 Negative 7 9 13 Weak 91 4 12 Strong The table totals numbers of results on the five duplicate pairs of monoclonals in Table II.
The reliability of the cell sonicate ELISA experiments is assessed in Table VI, which shows several large discrepancies within duplicate pairs. Finally, the reliability of each of the four types of experiment are summarised in Table VII which clearly shows the poorer reliability of immunocytochemistry data.
Comparing amounts and strengths of staining In both immunocytochemistry and immunohistology experiments, amounts and strengths of staining were recorded for each antibody-target combination. Amounts and strengths are compared for immunocytochemistry (Table VIII) and immunohistology (Table IX). A negative strength is uniquely associated with 0% amount, but there is no association between amount and strength in non-negative reactions. The revised target-cell list Table X lists the target-cell types used in the statistical analyses. This list was based on the target-cell types used by the data entry programs, and was augmented to include 'other' cell types used by participants. The list distinguishes cell types by type of experiment; for example, SCLC classic cell lines obtain three entries in the list: one for flow cytometry, another for ELISA and a third for immunocytochemistry. This distinction was made in anticipation of differences in reactivity between experiment types for a given antibody and cell type. For immunohistology, some cell types have been distinguished by the tissue on which they were found. For example
5
Table VI Reproducibility in Cell Sonicate ELISA experiments Mean absorbance on 2nd mab of pair" Mean absorbance >6.0 2.5-6.0
Br. J. Cancer
'."
3-9 63, XIV, 3 Suppi. XIV, 63, Suppl.
1991 Macmillan Press Ltd., 1991 Ltd.,
9
Data analysis of the Second International Workshop Cancer Antigens
on
Small Cell Lung
W.R. Gilks', L. Oldfield' & P. Wild2 'Medical Research Council Biostatistics Unit, Fair View Lodge, 5 Shaftesbury Road, Cambridge CB2 2BW, UK; 2Departement Environnement Chimique Service Epidemiologie, Avenue de Bourgogne, B.P. 27, 54501 Vandoeuvre Cedex, France. Summary Methods of data collection for the 2nd Small Cell Lung Cancer Workshop are described, and data reliability is reviewed. The method of cluster analysis of the workshop antibodies is described and discussed. Of the 27,1 11 results submitted 20,705 were judged to be reliable for analysis and 13,802 of these came from immunohistology experiments. Data derived from immunocytochemistry experiments were somewhat less reproducible than flow cytometry, immunohistology and ELISA experiments. The cluster analysis was developed from methods employed in the leucocyte antigens workshops. Several checks on the methods of cluster analysis and the transformation of data did not substantially alter the final groupings. The workshop confirms that, although there are some methodological difficulties, the cluster analysis can successfully be applied to data derived largely from immunohistology, and thus has applicability to other tumour types.
The main purpose of the data analysis of the Second International Workshop on Small Cell Lung Cancer (SCLC) Antigens was to cluster Workshop monoclonal antibodies into groups having similar patterns of reactivity. We report here on the methods of data collection and analysis and also present some summary results. More detailed reporting and discussion of results can be found elsewhere in this volume. Methods and results Data collection For the First SCLC Workshop, data were collected via paper forms. Following our initiative in the 4th Leucocyte Antigens Workshop (Gilks et al., 1989), participants in the Second Workshop were requested to submit their data on computer diskettes. To support this, data entry programs for use on IBM-PC microcomputers (or compatibles) were specially written, and sent to all participants. These programs allowed participants to enter, check, edit and if necessary delete their experimental data. Participants were also provided with paper forms to assist in the transfer of data from laboratory to computer. Three data entry programs were written: one for flow cytometry or fluorescence microscopy data; another for immuno-cytochemistry data, and the third for immunohistology data. The flow cytometry/fluorescence microscopy data entry program requested the following information for each experiment: target cell type (from a list of 20 normal and tumour cell types); species of secondary antibody (goat, rabbit, rat or sheep); and type of experiment (flow cytometry or fluorescence microscopy). For each antibody tested in each experiment, the data entry program requested the percentage of cells reacting (for flow cytometry, the percentage of cells fluorescing more intensely than a negative gate). Below, we refer to this as the amount of reactivity. The immunocytochemistry data entry program requested the following information for each experiment: target cell type (from the list referred to above); species of secondary antibody (as above); method of staining (indirect immuno-
peroxidase, PAP, APAAP, indirect immunofluorescence, immunogold or ABC); and method of fixation (none, acetone or formalin). For each antibody tested in the experiment the following were requested: the strength of staining (negative,
Correspondence: W.R. Gilks.
weak or strong); the amount of staining (0%, 1-25%, 26-75% or >75% of cells stained); and the location of staining (nucleus, cytoplasm, surface, or a mixture of these). The immunohistology data entry program requested the following information for each experiment: the type of tissue (from a list of 23 tissues); the species of secondary antibody (as above); the method of staining (as above); and the method of fixation (frozen - no fixation, frozen - brief acetone, formalin, or Bouin's). Also, the types of cell identified in the tissue were also requested (from a common list of 39 normal and tumour cell types). Then, for each cell type identifiable in the tissue, and for each antibody, the strength of staining (as above) and the amount of staining (as above) were
requested.
Where appropriate, each list of possible answers to each question included an 'other' option. Thus, for example, if a target cell used in a flow cytometry experiment was not included in the preset list of 20, the participant would be able to select the 'other' option and then record the full details. The items of data requested for the Second Workshop closely followed those of the First Workshop (Gilks et al., 1988), the major differences in the present workshop being that, for flow cytometry/fluorescence microscopy experiments no strength data were requested; for immunocytochemistry experiments the amount of staining was requested in the form of percentage bands and the location of staining was also
requested.
Collection of data via the data entry programs had the
advantage that participants were able to check and print out their own computerised data at source. It also allowed for rapid central data assimilation, shortening the time needed to prepare interim statistical analyses for feedback to participants. In the case of immunohistology, the data entry
program provided another important advantage. We found in the 4th Leucocyte Antigens Workshop that participants were reluctant to record negative reactions on all combinations of cell-type and antibody where they occurred, simply because of the multiplicity of such combinations in any immunohistology experiment. Instead, some participants recorded only reactions which were positive. This presented difficulties in analysing the data, as it was sometimes difficult to discern whether a cell-type did not react with any of the antibodies in the panel, or whether the cell-type was not present or not identified in the tissue sections. In the present Workshop, the immunohistology data entry program removed this problem by first asking the participant to state the cell-types identified in the tissue (as described above), and then forcing the participant to provide data for all these cell-types against each antibody tested. Clearly, enforcing such rigour in data recording would have been impossible using only paper data forms.
4
W.R. GILKS et al.
Data received For present purposes we define an experiment as the simultaneous testing of several antibodies. We define an individual experimental result as the data pertaining to one antibody in one experiment against one cell type. Table I summarises the experimental results recevied. The cell sonicate ELISA data was unsolicited, but was provided by one laboratory via the immunocytochemistry data entry program. These ELISA data derived from cytospins of cell lines which were grown in culture medium, harvested, washed, sonicated until 50-75% of the cells were broken, then used to coat plates in the presence of dilute glutaraldehyde, and then incubated with the primary antibodies, after which
peroxidase-conjugated bridging antibody and substrate were added and absorbances read at 492 nm. Several experiments did not include all Workshop antibodies: indeed, one laboratory included only four antibodies in each histology experiment. The amount of work done is more accurately reflected in the column headed 'Results contributed' in Table I. The numbers of results are roughly comparable with those of the First Workshop (Gilks et al., 1988). Data exclusions To ensure that data for analyses were reasonably reliable, experiments were excluded from analyses if negative controls in the antibody panel were found to react positively, or if there were large differences in reactivity amongst replicate antibodies (i.e. antibodies replicated in the Workshop panel). Thus the following groups of Workshop antibodies were used to determine data exclusions:
Antibody group Negative controls Replicate group 1 Replicate group 2 Replicate group 3
Workshop numbers 01,32,51,72 11,53 31,41,58,82 03,28,56,71
Antibody name PBS + 5% FCS MOC-31 NCC-LU-246
CAM5.2
The exclusion criteria were formulated as follows: (a) Flow cytometry: Exclude the experiment if either (i) > 10% amount of reactivity was recorded on any negative control or (ii) within two or more of the replicate groups, the amounts of reactivity on any pair of antibodies differed by more than 2.6 on the logistic scale. (The logistic transformation transforms a percentage y to the value lo&[y/ (100-y)]: this has the effect of stretching the ends of the percentage scale relative to the middle. For example, two replicate antibodies scoring 1% and 12% yield a difference of 2.6 on the logistic scale, as do scores of 12% and 65%) (b) Immunocytochemistry, Immunohistology and Cell Sonicate ELISA: Exclude the experiment if either (i) >0% amount of staining was recorded on any negative control antibody or (ii) within two or more of the replicate groups, 0% amount of staining was recorded on one antibody but not on another, or 1-25% was recorded on one and 76-100% was recorded on another. Table I summarises the experiments excluded, and the experimental results remaining.
Data reliability Tables II-VI demonstrate data reliability, by way of comparisons on five duplicate pairs of antibodies after making the exclusions detailed above. These five duplicate pairs were as follows, by Workshop number: 11 and 53 (both MOC-3 1); 31 and 41 (both NCC-LU-246); 58 and 82 (again both NCC-LU-246); 3 and 28 (both CAM5.2); and 56 and 71 (again both CAMS.2). To introduce the methodology, Table II compares antibodies 11 and 53 in Flow Cytometry experiments. To facilitate comparisons with Immunocytochemistry and Immunohistology data (Tables IV and V), amounts of reactivity in Table II have been grouped into ranges of percentages. Thus, for example, out of nine flow cytometry experiments in which antibody 11 scored in the range 1-25%, antibody 53 scored outside that range only once. As expected for a duplicate pair of antibodies, the results in Table II are concentrated on the
diagonal.
To summarise results on duplicate antibodies in flow cytometry experiments, Table III combines the results from Table II with similar results on the remaining four duplicate pairs of antibodies. Thus the 70 results in the second row and second column of Table III include the eight results from the corresponding cell of Table II. As expected, the results concentrate strongly on the diagonal of Table III, although the number of off-diagonal results (indicating lack of experimental precision) is perhaps surprising. In Table IVa, the reliability of the amount of staining in immunocytochemistry experiments is demonstrated as for flow cytometry experiments in Table III. Table IVa suggests that immunocytochemistry experiments are somewhat less reliable than flow cytometry experiments; (recall also from Table I that a high proportion of immunocytochemistry experiments were excluded before compiling Table IV). Table IVb presents a similar analysis based on the strength of staining in immunocytochemistry experiments, and demonstrates that strength of staining is no more or less reliable than amount of staining. Table Va demonstrates the reliability of the amount of staining in immunohistology experiments. It has been constructed similarly to Tables III and IVa, except that each result tabulated represents the reactions of a duplicate pair of antibodies in one cell-type in one experiment. Table Va shows that a negative reaction on one antibody is generally accompanied by a negative reaction on its duplicate. A positive (76-100%) reaction is about as reliable as for flow cytometry experiments. However, an intermediate (1-75%) reaction on one antibody bears no relationship at all with the reaction on its duplicate. Similar conclusions can be drawn from Table Vb, regarding the reliability of strength of staining data from immunohistology experiments. Table II Comparison of antibodies 11 and 53 (both MOC-31) in Flow Cytometry experiments Amount of reactivity Amount of reactivity on antibody 53 on antibody 11 0% 1-25% 26-75% 76-100% 0% 4 1 1 0 1-25% 0 8 1 0 26-75% 0 3 0 1 76-100% 0 0 0 9 The numbers in the body of the table are numbers of experiments.
Table I Data contributed to the 2nd SCLC Workshop Laboratories Experiments Results Experiments Results Experimental type contributing contributed contributed excluded remaining Flow cytometrya 7 49 3,643 6 3,006 Immunocytochemistry 4 34 2,329 11 1,405 Immunohistology 12 196 18,646 13 13,802 Cell sonicate ELISAb 1 26 2,493 0 2,493 All 305 20c 27,111 30 20,705 aNo Fluorescence Microscopy data were contributed. bExperimental data of this type wre not solicited. cThree laboratories contributed data from more than one type of experiment.
DATA ANALYSIS
Table III Reproducibility in Flow Cytometry experiments Amount of reactivity on second mab of pair Amount of reactivity 1-25% 26-75% 76-100% in first mab of pair 0% 1 2 2 21 0% 0 1 70 6 1-25% 2 3 1 15 26-75% 23 1 0 0 76-100% The table totals numbers of results on five pairs of duplicate monoclonals: (workshop numbers 11 and 53; 31 and 41; 58 and 82; 3 and 28; and 56 and 71). Table IV Reproducibility in Immunocytochemistry experiments (a) Amount of staining Amount of staining in 2nd mab of pair in 1st mab of pair 1-25% 26-75% 76-100% 0% 1 0 2 25 0% 3 3 1-25% 4 0 2 3 2 0 26-75% 21 2 0 0 76-100% Strength of staining in 2nd mab ofpair (b) Strength of staining Strong Weak in Ist mab of pair Negative 0 3 25 Negative 6 5 2 Weak 5 21 1 Strong The table totals numbers of results on the five duplicate pairs of monoclonals in Table II. Table V Reproducibility in Immunocytochemistry experiments Amount of staining in 2nd mab of pair (a) Amount of staining 1-25% 26-75% 76-100% 0% in Ist mab of pair 4 10 3 379 0% 1 0 9 6 1-25% 4 8 7 6 26-75% 8 85 1 1 76-100% Strength of staining in 2nd mab of pair (b) Strength of staining Weak in Ist mab of pair Strong Negative 6 10 379 Negative 7 9 13 Weak 91 4 12 Strong The table totals numbers of results on the five duplicate pairs of monoclonals in Table II.
The reliability of the cell sonicate ELISA experiments is assessed in Table VI, which shows several large discrepancies within duplicate pairs. Finally, the reliability of each of the four types of experiment are summarised in Table VII which clearly shows the poorer reliability of immunocytochemistry data.
Comparing amounts and strengths of staining In both immunocytochemistry and immunohistology experiments, amounts and strengths of staining were recorded for each antibody-target combination. Amounts and strengths are compared for immunocytochemistry (Table VIII) and immunohistology (Table IX). A negative strength is uniquely associated with 0% amount, but there is no association between amount and strength in non-negative reactions. The revised target-cell list Table X lists the target-cell types used in the statistical analyses. This list was based on the target-cell types used by the data entry programs, and was augmented to include 'other' cell types used by participants. The list distinguishes cell types by type of experiment; for example, SCLC classic cell lines obtain three entries in the list: one for flow cytometry, another for ELISA and a third for immunocytochemistry. This distinction was made in anticipation of differences in reactivity between experiment types for a given antibody and cell type. For immunohistology, some cell types have been distinguished by the tissue on which they were found. For example
5
Table VI Reproducibility in Cell Sonicate ELISA experiments Mean absorbance on 2nd mab of pair" Mean absorbance >6.0 2.5-6.0
